TUDP removes timestep conditioning from diffusion policies and adds an action-discrimination signal to learn a time-unified velocity field, achieving SOTA RLBench success rates (82.6% multi-view, 83.8% single-view) and stronger few-iteration performance than 3D Diffuser Actor.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Time-Unified Diffusion Policy with Action Discrimination for Robotic Manipulation
TUDP removes timestep conditioning from diffusion policies and adds an action-discrimination signal to learn a time-unified velocity field, achieving SOTA RLBench success rates (82.6% multi-view, 83.8% single-view) and stronger few-iteration performance than 3D Diffuser Actor.